The Implementation Of Continuous Wavelet Transform Using Switched-Current Circuit

The Wavelet Transform (WT) is the milestone of the developing history of Fourier Transform. As a new self-adapted time&frequence-domain analysis tool, it can tunes the time-frequency window automatically and has a strong flexibility. It has found a wide range of applications in the processing of signals due to its time-frequency localization characteristics. But in practice, Because of the heavy computational cost and too much time waste, software implementations of wavelet transform cannot be used in real-time signal processing. This effects the application of WT, so it calls for hardware implementation.The switched-current (SI) circuit is a new technology to process analog sample-data which compatible with standard CMOS technics. It has many advantages because current domain operation offers greater ease for low voltage operation and low power dissipation,and furthermore, the designing method is systematization and modularization, so the SI circuit fits to realize large-scale circuit. It does not require the linear capacitors, and the SI integrator does not require the voltage operation amplifier. It makes circuit simpler and reduces the harmful effects.In view of the current situation above, SI technology is a wise choose to implement WT. Therefore, in this paper, two methods of implementing continuous wavelet transform (CWT) using switched-current technology are proposed.One method is using leap-frog switched-current filters which has the advantage of the low sensitivity of pass band and insensitivity to parasitic capacitance. Simultaneously, it has mature theory of designing. This method can approximating CWT better and fit the demands of real-time signal processing.The other is based upon the frequency-domain designing method of WT implementation, in this paper, the method is improved. Trapezoid formula is used to approaching numerical integration, and according to this principle, a bank of bandpass filters was improved and that was implemented by biquadratic filters of SI circuit. The errors were detailed analysised and we can see the precision was heightened.Taking the implementation of several channels of Marr wavelet for example, the designing courses of the two methods were expounded. The simulation results indicate the feasibility of these improvings.